Pure Appl. Geophys.
168 (2011), 2097–2119
Ó 2011 Springer Basel AG (outside the USA)
DOI 10.1007/s00024-011-0291-5
Nearshore Tsunami Inundation Model Validation: Toward Sediment Transport Applications
ALEX APOTSOS,1 MARK BUCKLEY,2 GUY GELFENBAUM,1 BRUCE JAFFE,2 and DEEPAK VATVANI3
Abstract—Model predictions from a numerical model,
Delft3D, based on the nonlinear shallow water equations are
compared with analytical results and laboratory observations from
seven tsunami-like benchmark experiments, and with field obser-
vations from the 26 December 2004 Indian Ocean tsunami. The
model accurately predicts the magnitude and timing of the mea-
sured water levels and flow velocities, as well as the magnitude of
the maximum inundation distance and run-up, for both breaking
and non-breaking waves. The shock-capturing numerical scheme
employed describes well the total decrease in wave height due to
breaking, but does not reproduce the observed shoaling near the
break point. The maximum water levels observed onshore near
Kuala Meurisi, Sumatra, following the 26 December 2004 tsunami
are well predicted given the uncertainty in the model setup. The
good agreement between the model predictions and the analytical
results and observations demonstrates that the numerical solution
and wetting and drying methods employed are appropriate for
modeling tsunami inundation for breaking and non-breaking long
waves. Extension of the model to include sediment transport may
be appropriate for long, non-breaking tsunami waves. Using
available sediment transport formulations, the sediment deposit
thickness at Kuala Meurisi is predicted generally within a factor
of 2.
Key words: Tsunami, inundation, run-up, benchmark valida-
tion, numerical modeling.
1. Introduction
The 26 December 2004 tsunami in the Indian
Ocean killed over 237,000 people and left more than
a million homeless (U.S. Agency for International
Development (USAID), 2005). With a large segment
of the world’s population living along coasts, major
1
Pacific Coastal and Marine Science Center, USGS, 345
Middlefield Rd, MS 999, Menlo Park, CA, USA. E-mail:
aapotsos@usgs.gov
2
Pacific Coastal and Marine Science Center, USGS, 400
Natural Bridges Drive, Santa Cruz, CA, USA.
3
Deltares, Delft, The Netherlands.
Pure and Applied Geophysics
tsunamis have the potential to produce similar
disasters in the future. In order to develop appropriate
coastal management strategies that properly plan for
and mitigate losses owing to tsunamis, researchers
have generally focused on two approaches. The first
involves coupling early detection systems with
numerical models that predict, in real-time, areas
most likely to be impacted by a tsunami (e.g., WEI
et al., 2008; TANG et al., 2009). The second involves
improving local tsunami hazard assessments through
a better understanding of the recurrence interval and
magnitude of historic and pre-historic tsunamis (e.g.,
GONZÁLEZ et al., 2009). Both of these approaches
require a detailed understanding of the physical pro-
cesses that occur during a tsunami, particularly the
dynamics of nearshore propagation and inundation.
The second approach also requires a detailed under-
standing of how tsunamis transport and deposit
sediment. Here, we validate a hydrodynamic model
and demonstrate that extending it to include sediment
transport produces reasonable results that compare
well with field observations.
Unfortunately, the infrequent and unpredictable
nature of tsunamis makes it difficult to collect detailed
measurements of the physical processes that occur.
Therefore, to better understand tsunami processes,
analytical, laboratory, and numerical experiments have
been conducted. While much of the early work focused
on analytical solutions to simplified problems
(e.g., CARRIER and GREENSPAN, 1958; TADEPALLI and
SYNOLAKIS, 1994, 1996; CARRIER et al., 2003; KÁNOGLU,
2004) and laboratory experiments employing solitary
waves (e.g., SYNOLAKIS, 1987; BRIGGS et al., 1995,
1996), recent increases in processing power have
greatly improved the utility of numerical models.
These models can now be used to predict the propa-
gation and inundation paths of tsunamis in almost real-
time (e.g., TITOV et al., 2005; WEI et al., 2008; TANG
2098 A. Apotsos et al.
et al., 2009). The use of numerical models in this
operational sense has allowed for the rapid determi-
nation of which communities are at risk once a tsunami
has been detected, and has helped coastal planners and
emergency responders act accordingly.
Numerical models can also play a role in
improving tsunami hazard assessments, which are
often constrained by our limited knowledge of the
historic and pre-historic tsunami record. In order to
estimate the probability that a major tsunami will
strike a specific location, information is needed
concerning tsunamis that occurred hundreds or
thousands of years ago. Typically the only evidence
that remains from these paleo-tsunamis is preserved
in the geologic record as a sedimentary layer (e.g.,
ATWATER, 1987). Dating the soil layers surrounding
these deposits can help identify approximately when
the tsunami occurred, and thus increase our under-
standing of the recurrence interval of major tsunamis.
To improve hazard assessments, knowledge of the
hydrodynamic characteristics of the tsunami is also
necessary. Therefore, significant efforts have been
made to use the information (e.g., sediment size,
deposit thickness, grading) preserved in tsunami
sediment deposits to estimate the wave height, flow
depth, and velocity of paleo-tsunamis (e.g., JAFFE and
GELFENBAUM, 2007; SMITH et al., 2007; NANAYAMA
et al., 2007). These efforts, as well as efforts to
understand the scouring that occurs around obstacles
during tsunamis, would benefit from a better under-
standing of how sediment is transported and
deposited by a tsunami. Numerical models that
accurately predict the hydrodynamic processes of a
tsunami could be coupled with sediment suspension
and transport formulations and used to determine
which processes dominate tsunami sediment transport
and what information concerning the source mecha-
nism and hydrodynamic characteristics of the tsunami
is retained in tsunami sediment deposits.
Before a hydrodynamic model can be coupled
with sediment transport formulations, it must be
shown to depict accurately the hydrodynamic pro-
cesses of a tsunami. Recent deployment of DART
(Deep-Ocean Assessing and Reporting of Tsunamis)
underwater buoys (e.g., GONZALEZ et al., 2005) and
the use of satellite imagery (e.g., ARCAS and TITOV,
2006) have improved model validation in the deep
Pure Appl. Geophys.
ocean. However, while analyses of video recordings
have provided a few estimates of onshore tsunami
flow velocities (FRITZ et al., 2006), detailed obser-
vations in the nearshore and on land, where sediment
transport is important, are limited. Model evaluation
for these regions is therefore still primarily done
using analytical and laboratory experiments.
Over the past two decades attempts have been
made to standardize the process of model evaluation
to ensure the accuracy of model predictions. To this
end, a set of benchmarks was developed during the
International Workshops on Long-Wave Run-up
(e.g., YEH et al., 1996; LIU et al., 2008), and a pro-
cedure for model verification was outlined
(SYNOLAKIS et al., 2007, 2008). Following the estab-
lishment of these initial benchmarks, several other
useful laboratory experiments have been conducted
and the data made available (see BALDOCK et al.,
2008; SWILGER, 2009). However, while detailed
comparisons with laboratory observations remain a
widely used method to test and validate numerical
tsunami models, especially in the nearshore, the use
of laboratory observations is not without problems
(see MADSEN et al., 2008 and Sect. 4.1 in this paper),
and offers only qualified model validation.
Here the results from a hydrodynamic numerical
model, Delft3D, are compared with seven tsunami-
like benchmarks, and with maximum water level
observations measured near Kuala Meurisi, Sumatra,
following the 26 December 2004 Indian Ocean tsu-
nami (JAFFE et al., 2006). The benchmarks include:
(1) the analytical solution of run-up on a plane beach
(CARRIER et al., 2003), and laboratory experiments
using (2) a piece-wise linear bathymetry (BALDOCK
et al., 2008), (3) an alongshore non-uniform beach
(SWILGER, 2009), (4) a circular island amid a constant
depth basin (BRIGGS et al., 1995), (5) an alongshore
non-uniform beach with a circular island (http://
isec.nacse.org/workshop/isec_workshop_2009), (6) a
complex three-dimensional beach (Monai Valley,
Japan as described in SYNOLAKIS et al. (2008), and (7)
a piece-wise linear bathymetry ending in a vertical
wall (BRIGGS et al., 1996).
Our primary goal in this paper is to demonstrate,
to the extent possible, that Delft3D is appropriate for
modeling the hydrodynamic processes that occur
during the nearshore propagation and inundation of a
Vol. 168, (2011) Tsunami Model Validation: Toward Sediment Transport
tsunami, and to examine the appropriateness of
including sediment transport in the model. Verifica-
tion that Delft3D accurately predicts the
hydrodynamic processes that occur during inundation
is advantageous because the model includes a
sophisticated sediment transport model that has been
extensively validated in a wide variety of coastal
settings (LESSER et al., 2004; VAN RIJN et al., 2007).
This transport model allows sediment particles of
varying sizes to be transported in bed or suspended
load, the vertical and horizontal structures of any
onshore sediment deposits to be determined, and the
morphological change associated with sediment
transport to be predicted. These features position the
model well to be a powerful tool for improving our
understanding of how tsunamis transport and deposit
sediment.
The numerical model is described in Sect. 2 and
the model-data comparisons are outlined in the sub-
sections of Sect. 3. The relevance of the benchmarks
to real tsunamis, a preliminary application of the
model to sediment transport, and the applicability of
the model are discussed in Sect. 4. Conclusions are
given in Sect. 5.
2. Model
Delft3D is a three-dimensional numerical model
that can simulate coupled hydrodynamic/sediment
transport/morphological change processes. This study
examines primarily the hydrodynamic model, which
solves the nonlinear shallow water equations (NLS-
WEs) on a three-dimensional staggered grid using a
finite difference scheme (STELLING and VAN KESTER,
1994) and the Alternating Direction Implicit time-
integration method (STELLING and LEENDERTSE, 1991).
The Delft3D model can be run as either a depth-
averaged or vertically layered (i.e., three-dimen-
sional) model. In the horizontal, a curvilinear or
spherical staggered grid is used. In the vertical (in the
case of three-dimensional simulations) sigma layers,
which are a fixed percentage of the water depth, or
fixed elevation layers are used. When the model is
run in three dimensions, the vertical velocity within
each grid cell is calculated from the continuity
equation due to the hydrostatic assumption and
2099
incompressibility of the flow. The extension of the
model to three dimensions is important for many
applications where resolution of the bottom boundary
layer or vertical stratification within the flow is
important (i.e., see Sect. 4.2).
Numerous methods have been developed to solve
the NLSWEs. The numeric solver used here is based
on the conservation of mass, momentum (during flow
expansions), and energy head (during flow contrac-
tions) (STELLING and DUINMEIJER, 2003). This method,
referred to as the Flood Solver within the Delft3D
framework, was developed specifically to handle
rapidly varying flows, accurately predict the rapid
wetting and drying of grid cells, and be applicable to
a wide range of Froude numbers (including subcriti-
cal and supercritical flows). This numerical method
has been shown to model well a variety of dam break
flooding problems (STELLING and DUINMEIJER, 2003),
and the model formulation appears well suited for the
simulation of long tsunami waves.
While the non-conservative form of the NLSWEs
has no unique solution at local discontinuities, the use
of conservative properties, as is done here, is often
sufficient to provide solutions that are acceptable in
terms of the local energy losses in and the propaga-
tion speed of a bore. This is because the conservation
of mass and momentum should remain valid even for
discontinuities in rapidly varying flows (i.e., breaking
tsunami waves) (ZIJLEMA and STELLING, 2008), and
because the dissipation of energy associated with
wave-breaking-generated turbulence is inherently
accounted for if momentum is conserved (HIBBERD
and PEREGRINE, 1979; BROCCHINI and PEREGRINE,
1996). However, the assumption of hydrostatic pres-
sure (i.e., a basic assumption of the NLSWEs) is
violated near the break point where vertical acceler-
ations within the wave are necessary to balance the
steepening of the wave front. Therefore, models
based on the NLSWEs will not capture accurately the
internal wave dynamics near the point of wave
breaking. However, once breaking has started, wave
front steepening is balanced by wave-breaking-gen-
erated turbulence (SVENDSEN and MADSEN, 1984), and
the NLSWEs should be valid for predicting the run-
up of broken waves (ZIJLEMA and STELLING, 2008).
A variety of methods are available to model the
rapid wetting and drying of grid cells (i.e., the
2100 A. Apotsos et al.
moving shoreline) (see ZIJLEMA and STELLING, 2008
for a brief discussion of the methods available). Here
a simple method that does not require special drying
and flooding procedures is used (STELLING and DUIN-
MEIJER, 2003), which ensures that the water depth is
always positive and requires that only one grid cell
is flooded or dried per time step. A single time step is
chosen such that this criterion is met throughout each
model simulation.
3. Numerical Experiments
The primary goal of this paper is to demonstrate
the general applicability of the model for simulat-
ing tsunami inundation. The model is, therefore,
compared with a wide array of different experi-
ments that include several different types of initial
wave forms, a wide range of initial nonlinearities,
two- and three-dimensional basins, and breaking
and non-breaking waves (Table 1). The input
parameters used are not tuned to the observations,
but instead physically realistic values were selected
a priori. Owing to the limitations of using labora-
tory experiments and analytical solutions to
evaluate tsunami inundation models and of using
the NLSWEs to model solitary waves, the model-
data comparisons focus on determining if the model
is generally appropriate for simulating tsunamis
(i.e., the model predictions are within *10% of the
observations for a variety of numerical and ana-
lytical benchmarks).
For the seven benchmark experiments, the model
is run as a depth-averaged model. Friction is param-
eterized using Manning’s formulation, with
n = 0.0001 for the analytical solution where friction
is neglected, and n = 0.01, a value consistent with
smooth concrete (ARCEMENT and SCHNEIDER, 1990),
for the six laboratory experiments. For each bench-
mark, an appropriate time step and grid cell size were
selected to ensure all important wave and bathymetric
features were resolved, the stability criterion in the
model that prevents more than one grid cell from
becoming wet/dry during a single time step was sat-
isfied (i.e., (Dt|u|)/Dx \ 2, where Dt is the time step,
u is the flow velocity and Dx is the grid cell size), and
the numerical experiments were simulated in an
Pure Appl. Geophys.
efficient manner (i.e., the largest time step that
resulted in no significant change in the results was
used). Unless otherwise specified, the cross-shore
location (x) is defined as positive onshore from the
initial shoreline, the alongshore coordinate (y) is
positive to the left of the centerline looking onshore,
the elevation (h) is positive up from the still water
level (SWL), and d is the water depth.
3.1. Run-up on a Planar Beach
3.1.1 Setup
The propagation of an initial free surface disturbance
composed of a leading depression followed by a
smaller amplitude crest (Fig. 1a, solid grey curve)
over an idealized planar beach with a uniform slope
of 1:10 (Fig. 1b, solid black line) was solved
analytically using the initial-value-problem (IVP)
technique (CARRIER et al., 2003) and the solutions
were provided by the organizers of the Third
International Workshop on Long-Wave Run-up Mod-
els (LIU et al., 2008). Cross-shore snapshots of the
free surface elevation (Fig. 2a, solid black curves)
and depth-averaged velocity (Fig. 2b, solid black
curves) at t = 160, 175, and 220 s, representing the
initial rundown, maximum rundown, and maximum
run-up, respectively, were provided. Time series of
the shoreline elevation (Fig. 2c, solid black curve)
and velocity (Fig. 2d, solid black curve) from
t = 100 to 280 s were also provided.
3.1.2 Results
Model predictions of the cross-shore water levels and
depth-averaged velocities agree well with the analyt-
ical solution at all three time steps (Fig. 2a, b
compare solid black and dashed grey curves). The
predicted depth-averaged velocity increases sharply
near the shoreline for t = 160 and 175 s, possibly
owing to the depth dependence of the Manning
friction formulation. However, this discrepancy is
confined to water depths less than 0.2 m, and does not
affect the model-data agreement in deeper water. The
predicted shoreline elevation (Fig. 2c) and velocity
(Fig. 2d) are in excellent agreement with the analyt-
ical solution.
Vol. 168, (2011) Tsunami Model Validation: Toward Sediment Transport 2101

Table 1
Benchmark summary

Benchmark Benchmark Wave form Initial wave Offshore Initial

type height (m) water depth (m) nonlinearity

(1) Planar beach Analytical Leading depression *11 5,000 0.002

(2) Piece-wise linear beach 2-D laboratory Solitary, impulse 0.06–0.54 0.51 0.12–1.06
(3) Alongshore non-uniform beach 3-D laboratory Solitary 0.39 0.78 0.5
(4) Constant depth basin with conical island 3-D laboratory Solitary 0.016, 0.032, 0.064 0.32 0.05, 0.10, 0.20
(5) Alongshore non-uniform beach 3-D laboratory Solitary 0.39 0.78 0.5
with conical island
(6) Complex 3-D bathymetry 3-D laboratory Leading depression *0.025 0.135 0.18
(7) Piece-wise linear beach with vertical wall 2-D laboratory Solitary 0.011, 0.065, 0.153 0.218 0.05, 0.30, 0.70

(a)
5
Water Level (m)

−5

−10

(b)
0
Elevation (m)

−2000

−4000

−50,000 −40,000 −30,000 −20,000 −10,000 0

Distance from Shoreline (m)

Figure 1
a Initial water level disturbance (solid grey curve) and still water level (dashed grey line) and (b) planar bathymetry with a slope of 1:10
(black curve) and initial water level (grey curve)

3.2. Run-up on a Piece-wise Planar Beach
3.2.1 Setup
A laboratory experiment using a two-dimensional
piece-wise linear bathymetry was conducted in
the large wave basin at Oregon State University
(BALDOCK et al., 2008) (Fig. 3). Ten different positive
impulse or solitary waves (Table 2, Trial01–Trial10)
were generated in 0.51 m water depth. The simulated
waves range in nonlinearity (e = Hi/d, where Hi is
the initial wave height) from 0.12 to 1.06, and include
bores, shore breaks, and a non-breaking wave. Water
levels were measured using 4 wire resistance wave
gauges and 3 ultrasonic wave gauges. Flow velocities
were measured using 5 Nortek acoustic doppler
velocimeters (ADVs). Two wire resistance wave
gauges and one ADV were located near the limit of
the wave maker during all runs (e.g., Fig. 3, vertical
hash labeled offshore) to ensure the repeatability and
alongshore uniformity of the initial wave. The other
wave gauges and ADVs were affixed to a movable
bridge located at x = -3.00, -1.50, -0.75, 0.00,
0.25, 0.50, 1.00, 2.00, 3.00 and 6.00 m, respectively
(Fig. 3, vertical hashes labeled bridge locations). All
instruments were within *0.5 m of the bridge
location. Maximum inundation distances were
recorded visually for all trials. The raw water level
and velocity data were processed to remove spurious
data (see BALDOCK et al., 2008). However, as this data
set has not been used extensively, the data were
assessed visually, and time series containing signif-
icant noise or non-zero initial water levels were not
2102 A. Apotsos et al. Pure Appl. Geophys.

(a) (c) 20

Water Level (m)

20

Run−up (m)
t = 220 s 10
10
0
0
t = 175 s
−10 −10
t = 160 s
−20 −20
−800 −600 −400 −200 0 200 150 200 250 300

(b)
20 (d) 20
Depth−Averaged

Velocity (m/s)
Velocity (m/s)

t = 175 s 10

Shoreline
10
t = 220 s 0
0
−10
t = 160 s
−10 −20
−800 −600 −400 −200 0 200 150 200 250 300
Distance from Shoreline (m) Time (s)

Figure 2
a Water levels and (b) depth-averaged velocities at t = 160, 175, and 220 s and shoreline (c) elevation (i.e., run-up) and (d) velocity from the
model predictions (dashed grey curves) and the analytical solution (solid black curves). The thick black line in a and b is the bed elevation

0.6
Elevation (m)

0.4 Bridge Locations

Offshore
0.2
d=0.51 m
0
−0.2 1/30
−0.4 1/15

−20 −15 −10 −5 0 5 10 15 20 25

Distance from Shoreline (m)

Figure 3
Model setup for the piece-wise planar beach. Vertical hashes indicate sensor or bridge locations

used. For further description of the model setup and likely because the observed inundation distance,
initial analysis of the results see BALDOCK et al. 15.6 m, occurs at the junction of the onshore slope
(2008). and a horizontal section of the bathymetry (Fig. 3,
x = 15.6 m). A small overestimation of the water
3.2.2 Results velocity in shallow water will cause the wave to
overtop this transition, and flow along the horizontal
The model predicts well (i.e., typically within 10%) bed, resulting in a large change in the predicted
the observed maximum inundation distance for 9 of inundation distance. For both the observations and
the 10 trials (Fig. 4). The model predicts most model predictions the ratio of the maximum run-up
accurately the inundation distance for waves with elevation (R) to the initial wave height (Hi) is largest
small initial nonlinearity, where dispersive effects are for the non-breaking wave, and decreases with
small. While the difference between the model increasing Hi (Table 2), suggesting that wave break-
predictions and observations generally increases with ing reduces the inundation distance of a tsunami.
increasing inundation distance, the model does not The model predicts well the observed wave height
appear to be biased toward over- or underpredicting (within *10%) and wave arrival time (within 1 s) at
the observations. The model significantly overpre- most locations (e.g., Figs. 5 and 6), with the best
dicts the observed inundation distance for Trial01, model-data agreement found for the non-breaking
Vol. 168, (2011) Tsunami Model Validation: Toward Sediment Transport 2103

Table 2
Summary of initial wave conditions for the piece-wise linear beach. Table based on Table 3 in BALDOCK et al. (2008)

Trial # Wave type Hi (m) Surf condition Inundation R (m) Hi/d R/Hi
distance (m)

01 Impulse 0.55 Developed bore 15.6 0.52 1.06 0.95

02 Impulse 0.30 Bore 14.61 0.49 0.58 1.62
03 Impulse 0.17 Shore break 11.35 0.38 0.33 2.23
04 Impulse 0.13 Incipient breaker 10.46 0.35 0.25 2.68
05 Impulse 0.06 Non-breaking 5.97 0.20 0.12 3.32
06 Impulse 0.31 Shore break 11.48 0.38 0.60 1.23
07 Impulse 0.14 Shore break 8.34 0.28 0.27 1.99
08 Solitary 0.1 Incipient breaker 5.78 0.19 0.19 1.93
09 Solitary 0.25 Shore break 9.77 0.33 0.38 1.30
10 Solitary 0.35 Bore 11.83 0.39 0.67 1.13

wave (Fig. 5, Trial05), which has the smallest initial
nonlinearity. As the model does not include disper-
sion, it does not reproduce accurately the wave shape
at the offshore locations for highly nonlinear,
dispersive waves (i.e., large e). For these waves, the
model predicts that the wave steepens and breaks
further offshore than observed. However, the lack of
dispersion in the model does not significantly affect
the predicted timing of the initial wave, even when
the wave has broken offshore of the measurement
location (i.e., Fig. 5, Trial01 and Trial10).
The model predicts well the wave magnitude in
shallow water and on land, and thus the total decrease
in the wave height across the instrumented zone
(Fig. 6d, Table 3). However, the model does not
predict the observed shoaling near the break point
Predicted Maximum Inundation
(e.g., Fig. 5, Trial10 and Fig. 6b) or the initial spike
observed in the water levels during breaking (e.g.,
Fig. 5b, c, Trial10), though the model more accu-
rately predicts the bore height that follows this spike.
The good model-data agreement in terms of water
levels for Trial01, which is highly nonlinear, at all
bridge locations (e.g., Fig. 5, Trial01) is likely
because all measurements were taken well onshore
of the initial break point. The predicted dissipation in
wave energy is owing to the shock-capturing numer-
ical solution scheme used, and not bottom friction.
For example, using very low values of n (i.e., a
virtually frictionless model) does not affect the
predicted decrease in wave height.
The model predictions compare well with the
measured velocities for most trials and locations (e.g.,
Fig. 7). In shallow water the velocity is well
predicted and increases rapidly as the wave front
passes and then slowly decelerates until the flow is

16

14 directed offshore (Fig. 7c, d). In deep water the

velocity is less peaked and not as well predicted for
12
Distance (m)

highly nonlinear waves (Fig. 7a, b). Following run-

10
down, differences occur in the observed and
8 predicted time at which sensors become dry. How-
6 ever, this feature is highly sensitive to the value of
4 n used.
2

0
0 2 4 6 8 10 12 14 16 3.3. Run-up on an Alongshore Non-Uniform Beach
Observed Maximum Inundation Distance (m)
3.3.1 Setup
Figure 4
Maximum inundation distance from the observations and model
predictions for Trial02–Trial10. Dashed grey line is perfect
An experimental setup designed to represent a steep,
agreement alongshore, non-uniform continental slope onshore of
2104 A. Apotsos et al. Pure Appl. Geophys.

0.30
0.20 Trial10 Trial01 (a) Bridge at x = −3 m
0.10 Trial05
0
5 10 15 20 25 30 35 40 45 50

0.30
Water Level (m)
0.20 Trial10
Trial01
Trial05
(b) Bridge at x = 0 m
0.10
0
5 10 15 20 25 30 35 40 45 50

0.30
0.20 Trial01 (c) Bridge at x = 1 m
Trial10 Trial05
0.10
0
5 10 15 20 25 30 35 40 45 50

0.30
0.20 (d) Bridge at x = 3 m
Trial10 Trial01
0.10 Trial05
0
5 10 15 20 25 30 35 40 45 50
Time (s)

Figure 5
Water levels for Trial01, Trial05, and Trial10 measured using a resistance wave gauge (solid black curves) and predicted from the model
(dashed grey curves) with the bridge located (a) -3 m, (b) 0 m, (c) 1 m, and (d) 3 m from the initial shoreline

0.6 0.4
(a) x = −17.5 m (b) x = −1.5 m
0.3
0.4
Maximum Predicted Wave Height (m)

0.2

0.2
0.1

0 0
0 0.2 0.4 0.6 0 0.1 0.2 0.3 0.4

0.2
(c) x = 1 m (d) x = 3 m
0.1

0.08

0.1 0.06

0.04

0.02

0 0
0 0.05 0.1 0.15 0.2 0 0.05 0.1
Maximum Observed Wave Height (m)

Figure 6
Maximum predicted and observed wave heights at four cross-shore locations for the resistance wire gauges (squares and diamonds) and sonic
wave gauges (open and filled circles). Dashed grey line is perfect agreement, and positive x is shoreward of the initial shoreline

a deep, constant depth ocean was constructed in the wave 0.39 m high was generated in 0.78 m water
large wave basin at Oregon State University (see depth and allowed to propagate shoreward. Owing to
SWILGER, 2009 for more details) (Fig. 8a). A solitary its large amplitude and high nonlinearity (e = 0.5),
Vol. 168, (2011) Tsunami Model Validation: Toward Sediment Transport 2105

Table 3
Initial wave heights (Hi) and the wave height at x = 3 m for the observations (H3m,obs) and the model predictions (H3m,pred) from the piece-
wise linear beach

Trial 01 02 03 04 05 06 07 08 09 10

Hi (m) 0.55 0.30 0.17 0.13 0.06 0.31 0.14 0.10 0.25 0.35
H3m,obs (m) 0.078 0.08 0.065 0.064 0.047 0.046 0.037 0.029 0.035 0.046
H3m,pred (m) 0.085 0.08 0.072 0.067 0.05 0.046 0.041 0.023 0.039 0.046

2 Trial06 (a) Bridge at x = −3 m

Trial05
0

−2
20 25 30 35 40
Cross−shore Velocity (m/s)

2 Trial06 (b) Bridge at x = 0 m

Trial05
0

−2
20 25 30 35 40

2 Trial06 Trial05 (c) Bridge at x = 1 m

0

−2
20 25 30 35 40

2 Trial06 Trial05 (d) Bridge at x = 3 m

0

−2
20 25 30 35 40
Time (s)

Figure 7
Cross-shore velocities during Trial05 and Trial06 from the observations (solid black curves) and model predictions (dashed grey curves) for
the bridge located (a) -3 m, (b) 0 m, (c) 1 m, and (d) 3 m from the initial shoreline. The oscillations in the observed data occur because the
sensors did not always obtain an accurate measurement, especially during the passage of a breaking wave

the solitary wave broke on the steep shelf slope. 3.3.2 Results
Water levels and velocities were measured at 17 and
3 locations within the basin, respectively (Fig. 8a, The model predicts well (i.e., within 10%) the mag-
white circles). All observations were taken offshore nitude and timing of the initial wave in the constant
of the initial shoreline, and no inundation data were depth portion of the basin (Fig. 9a–b, compare solid
provided. These observations were part of a larger and dashed curves). However, similar to what was
experiment (see SWILGER, 2009), and the data used observed for the piece-wise linear beach, the model
here were provided as part of the ISEC Community does not accurately reproduce the observed wave shape
Workshop: Simulation and Large-Scale Testing of at these locations. As the wave propagates over the
Nearshore Wave Dynamics (http://isec.nacse.org/ steep shelf slope, it becomes more nonlinear and less
workshop/isec_workshop_2009/). dispersive, and the model predicts well the observed
2106 A. Apotsos et al. Pure Appl. Geophys.

Elevation (m)
(a) (b)

Alongshore Distance (m)

10 10 0.2

5 5 0

0 0 −0.2

−0.4
−5 −5

−0.6
−10 −10

−0.8
−20 −10 0 10 −20 −10 0 10
Distance from Shoreline (m)

Figure 8
Model setup for (a) the alongshore non-uniform beach and (b) the alongshore non-uniform beach with a conical island. Open white circles are
the locations where water level measurements are available, and solid white circles are where water level and velocity measurements are
available. In b, only velocity was measured at x = -5 m, y = -5 m. Depth contours are in 0.10 m intervals from -0.80 to 0.40 m. The small
differences observed in the depth contours between (a) and (b) are owing to the fact that the bathymetry provided in (a) is based on a lidar
survey, while the bathymetry in (b) is based on the idealized experimental setup (i.e., no lidar was available)

timing, magnitude and shape of the initial wave
(Fig. 9c–d). While the model predicts that the wave
height decreases shoreward owing to breaking, at the
most shoreward locations the maximum predicted
wave heights are larger (by approximately 40%) and
behind (by approximately 0.5 s) the observations
(Fig. 9e–f). It is unclear why the model fails to predict
the total decrease in wave height observed in this
experiment given the good agreement found for the
piece-wise linear beach. The discrepancies observed
here may be owing to the steep slope of the beach and
the proximity to wave breaking. Better model-data
agreement might be expected further shoreward and on
land where no observations were taken.
The model predicts well the large cross-shore
velocities associated with the passage of the initial
wave (e.g., Fig. 10a), but does not predict as accu-
rately the smaller velocities associated with
secondary and reflected waves. Away from the
centerline, the model predicts the magnitude of the
observed alongshore velocity, but the predictions lead
the observations for the largest secondary wave
(Fig. 10b, t * 15–20 s). This difference in the
velocities is not related to bed friction (i.e., a larger
n does not improve model-data comparisons), and
may indicate the model overpredicts the wave speed
in the very shallow water onshore of the instruments.
3.4. Run-up on a Circular Island
3.4.1 Setup
A conical island was constructed in the center of a
constant depth basin at the US Army Corps of
Engineers Waterways Experiment Station, Coastal
Engineering Research Center (LIU et al., 1995;
BRIGGS et al., 1995; KÁNOGLU and SYNOLAKIS, 1998)
(Fig. 11). The island had a base diameter of 7.2 m
with a slope of 1:4 rising to a height of 0.625 m
(0.305 m above SWL). As the shoreline is circular,
the cross-shore coordinate (x) is defined as positive
onshore from the wave generating offshore boundary.
Three solitary waves with e = 0.05, 0.1, and 0.2 were
generated in 0.32 m water depth and propagated
through the basin. Water levels and run-up are
compared at 4 (Fig. 11, white circles) and 9 locations,
respectively. The values for the run-up were obtained
from the benchmark website (http://chl.erdc.usace.
army.mil/).
3.4.2 Results
Consistent with the experiment (LIU et al., 1995), the
model predicts the wave refracts around the island.
The two refracted wave fronts collide in the lee of the
island, driving flow up the back of the island (i.e., in
Vol. 168, (2011) Tsunami Model Validation: Toward Sediment Transport 2107

0.40
0.20
(a) x = −18.5 m
0

0.40
0.20
(b) x = −14.5 m
0
Water Level (m)

0.40
0.20
(c) x = −13 m
0

0.40
0.20 (d) x = −11 m
0

0.40
0.20
(e) x = −9 m
0

0.40
0.20 (f) x = −5 m
0
0 5 10 15 20 25 30 35 40
Time (s)

Figure 9
Water levels at 6 cross-shore locations along the centerline of the basin for the observations (solid black curves) and model predictions
(dashed grey curves)

(a)
Velocity (cm/s)

150
Cross−shore

100
50
0
−50
0 5 10 15 20 25 30
(b)
Velocity (cm/s)
Alongshore

20

−20

0 5 10 15 20 25 30
Time (s)

Figure 10
(a) Cross-shore and (b) alongshore velocities from the observations (solid black curves) and model predictions (dashed grey curves) for
x = -5 m and y = -5 m

the direction opposite to initial wave propagation).
The model predicts well the timing and magnitude of
the observed water levels at the two locations in front
of the island and the location to the side of the island
(not shown). However, for large e, the front of the
predicted wave is steeper than the observations and
only for e = 0.05 does the predicted wave shape
match the observations. Behind the island, the model
predicts well the timing of the observed increase in
the water level owing to the arrival of the refracted
waves, but underpredicts the magnitude by about 20,
30 and 50% for e = 0.05, 0.1, and 0.2, respectively.
2108 A. Apotsos et al. Pure Appl. Geophys.

Elevation (m)
(a) 15 0.3

Alongshore Distance (m)

10 0.2

5 0.1

0 0

−0.1
−5

−0.2
−10

−0.3
5 10 15 20 25

(b)
Elevation (m)

0.2
d=0.32 m
0
−0.2
0 5 10 15 20 25
Cross−shore Distance (m)

Figure 11
Model setup for the circular island amid a constant depth basin

The 50% underprediction for e = 0.2 is similar to
that found following TITOV and SYNOLAKIS (1997) (as
shown in KÁNOGLU and SYNOLAKIS (1998)) but slightly
larger than for the analytical solution presented by
KÁNOGLU and SYNOLAKIS (1998).
The model predicts well the variation of run-up
around the island for all three waves (Fig. 12). The
model tends to overpredict the observations for small
e (Fig. 12a) and underpredict the observations for
large e (Fig. 12c). The former may be owing to the
predicted oversteepening of the wave in shallow
water due to the lack of dispersion in the model,
while the latter is likely caused by the predicted
waves breaking offshore of the island.
3.5. Run-up on an Alongshore Non-Uniform Beach
with a Conical Island
3.5.1 Setup
The underlying bathymetry in this experiment is
identical to that described in Sect. 3.3.1 for the
alongshore non-uniform beach, except that a conical
island was added at the apex of the steep shelf slope
(Fig. 8b). A solitary wave 0.39 m high was generated
in 0.78 m water depth and breaks on the steep shelf
slope. Water levels and velocities were measured at 9
and 3 locations, respectively (Fig. 8b, white circles).
Owing to the formation of a strong wake immediately
behind the island, the sensor located here did not
accurately record the water level or flow velocity (Pat
Lynett, pers. comm.), and these measurements are not
used. These data were provided as part of the ISEC
Community Workshop: Simulation and Large-Scale
Testing of Nearshore Wave Dynamics (http://isec.
nacse.org/workshop/isec_workshop_2009).
3.5.2 Results
The results for the measured water levels and
velocities are similar to those presented in Sect.
3.3.2, and are not discussed here. Instead compar-
isons are made with two features created by the
addition of the island. These include the formation
of a wake in the lee of the island following the
passage of the initial wave (Fig. 13, top panels) and
the convergence of four secondary waves slightly
behind and to the side of the island (Fig. 13, bottom
panels). Owing to the limited observational data
available at these locations, the model results are
Vol. 168, (2011) Tsunami Model Validation: Toward Sediment Transport
4 (a) ε = 0.05
0
0 0.5
4 (b) ε = 0.1
R/A
0
0 0.5
4 (c) ε = 0.2
0
0 0.5
Angle from Front of Island (Radians)
Normalized run-up (R/A), where A is the wave amplitude, for the observations (black diamonds) and model predictions (grey circles) for
e = (a) 0.05, (b) 0.1, and (c) 0.2
compared qualitatively with snapshots taken from
video recordings of the experiments (Pat Lynett,
pers comm.).
The model predicts the formation of a wake
behind the island, but the shape differs somewhat
from the observations (Fig. 13, compare top right and
left panels). Similarly, the model predicts the con-
vergence of four secondary waves, but the magnitude,
timing, and location of each wave differs somewhat
from the observations (Fig. 13, compare bottom right
and left panels). The slight misalignments of the four
waves likely account for the failure of the model to
capture the signal of this convergence in the water
level and the velocity time series (not shown) at the
single observation location (Fig. 13, black circle in
the bottom left panel).
3.6. Run-up on a Complex 3-D Beach
3.6.1 Setup
A 1:400 scale laboratory experiment of the impact of
the 1993 Okushiri tsunami on Monai, Okushiri
Island, Japan, including realistic bathymetry and
topography scaled to the tank, was conducted in
The Central Research Institute for Electric Power
Industry (CRIEPI) in Abiko, Japan (LIU et al., 2008)
(Fig. 14a). The experiment was forced with a leading
2109
1 1.5 2 2.5 3
1 1.5 2 2.5 3
1 1.5 2 2.5 3
Figure 12
depression N-wave at the offshore model boundary in
0.135 m water depth (Fig. 14b). Water level data
were provided at 3 stations inshore of the offshore
island (Fig. 14a). Owing to the complex shoreline,
the cross-shore coordinate (x) for this experiment is
positive shoreward from the wave generating off-
shore boundary, and the alongshore coordinate (y) is
positive from the right side of the experimental basin
facing onshore.
3.6.2 Results
The model predictions are in excellent agreement
with the observations at all three stations for the
initial wave (Fig. 15). The wave crest arrives at
approximately t = 15 s, with the model predictions
lagging the observations by less than 1 s and the
wave magnitude predicted within 10%. After the
first wave, the observed water levels are highly
influenced by reflections off the topography and
closed boundaries. The model predicts the arrival of
the first reflected wave at approximately t = 35 s,
but the predictions both lead and are larger than the
observations. While the model predicts the arrival of
a series of secondary waves at all three locations,
the accuracy of the predictions of these reflected
waves decreases with increasing time (Fig. 15,
t [ 55 s).
2110 A. Apotsos et al. Pure Appl. Geophys.

Figure 13
Snapshots of the water surface just after the passage of the initial wave (top left panel) and just as four secondary waves are about to collide
(bottom left panel) and the predicted water levels (color contours) and velocities (vectors) for similar times (right panels). The black circles
and the solid curves in the right panels are the location of an ADV and the depth contours, respectively. Note these panels are rotated 90°
counterclockwise with regards to Fig. 8b

3.7. Run-up on a Vertical Wall near the wall and between gauges 7 and 8, respec-
tively. The wave did not break for e = 0.05.
3.7.1 Setup

A piece-wise linear bathymetry composed of slopes 3.7.2 Results

of 1:53, 1:150, and 1:13 and ending in a vertical wall
was constructed in a 23.2 m long, 0.45 m wide flume
(BRIGGS et al., 1996) (Fig. 16). Three solitary waves
with e = 0.05, 0.3, and 0.7 were generated in
0.218 m water depth. Water levels are compared at
6 locations (Fig. 16). The maximum run-up was
recorded as the highest point water reached on the
vertical wall. For e = 0.3 and 0.7, the waves broke
For e = 0.05 (case A), the model predicts well the
timing, magnitude, and shape of the incoming wave
(e.g., Fig. 17a–c). The timing and magnitude of the
reflected wave are well predicted by the model, but
the predicted wave has a steeper vertical face than
observed (Fig. 17a, b). This discrepancy in the
reflected wave shape is because the predicted wave
Vol. 168, (2011) Tsunami Model Validation: Toward Sediment Transport 2111

Elevation (m)
(a) (b) 0.02
0.1
3

Offshore Water Level (m)

Alongshore Distance (m)
0.015

2.5 0.05
0.01
st9
2 0.005
0
st7
1.5 0
st5 −0.05
1 −0.005

0.5 −0.1 −0.01

−0.015
1 2 3 4 5 0 10 20 30
Cross−shore Distance (m) Time (s)

Figure 14
a Model setup and (b) water level at the offshore boundary for the complex 3-D beach. The location of the three water level stations are
labeled st5, st7, and st9 in (a)

0.04
(a) Station 5
0.02
0

10 20 30 40 50 60 70 80 90 100
Water Level (m)

0.04
(b) Station 7
0.02
0
10 20 30 40 50 60 70 80 90 100

(c) Station 9
0.04
0.02
0
10 20 30 40 50 60 70 80 90 100
Time (s)

Figure 15
Water levels at the three stations for the observation (solid black curves) and the model predictions (dashed grey curves)

0.4
Elevation (m)

G10
G7 G8 G9
0.2 G6
G5

d=0.218 m
0

−0.2
−20 −15 −10 −5 0
Distance to Vertical Wall (m)

Figure 16
Model setup for the piece-wise linear beach ending in a vertical wall
2112 A. Apotsos et al. Pure Appl. Geophys.

steepens in the shallow water before reaching the
wall, and maintains this steepened shape after
reflecting seaward. Conversely, the inherent disper-
sion of the solitary wave keeps the observed wave
from steepening both before and after reflection. The
predicted maximum run-up of 0.0246 m is within
approximately 10% of the observed maximum of
0.0274 m.
For e = 0.3 and 0.7 (cases B and C, respectively),
dispersion becomes more important, and the model
predicts well the timing, amplitude, and shape of the
incoming wave only at station 5 (Fig. 17d, g). Further
shoreward the model predicts that the wave steepens
and breaks at a location more offshore than observed.
Significant differences exist between the modeled and
observed reflected waves in terms of both the wave
magnitude and shape, likely owing to the way a
broken or breaking wave interacts with a vertical
surface. While the timing of the reflected wave is
generally well predicted for e = 0.3, for e = 0.7 the
predictions lead the observations by about 1 s at the
most seaward location (Fig. 17g). The predicted
maximum run-up values of 0.0936 m and 0.154 m
for e = 0.3 and 0.7, respectively, are about 4 and 2
times smaller than the observed values of 0.4572 m
and 0.2743 m. This underprediction is likely owing to
the measured run-up being a result of wave-breaking-
generated spray, which the model cannot predict, as
well as other local breaking processes that are not
included in the model. Differences in the wave shape
as it reaches the wall owing to a lack of dispersion in
the model may also contribute to the observed
underprediction.
3.8. Kuala Meurisi, Sumatra
3.8.1 Setup
Bathymetry, topography, and maximum water level
estimates were collected following the 26 December
2004 Indian Ocean tsunami near Kuala Meurisi on
the north coast of Sumatra, Indonesia (JAFFE et al.,
2006; GELFENBAUM et al., 2007) (Fig. 18a). The
maximum water levels were mostly estimated from
broken branches, and may represent minimum values
of the actual water levels. The bathymetry was
measured along seven cross-shore transects, while
topography was measured only along a single tran-
sect. As the bathymetry is approximately alongshore
uniform, a one-dimensional, depth-averaged model
with a cross-shore grid spacing of 12 m is used. The
offshore water level variation was obtained from the
output of a deep water ocean propagation model and
an estimate of the initial source disturbance for the 26

Case A Case B Case C

(a) Station 5 (g) Station 5
0.015
0.08 (d) Station 5
0.06 0.1
0.01 0.04
0.005 0.02 0.05
0 0 0
5 10 15 20 5 10 15 20 5 10 15
Water Level (m)

(b) Station 8 0.08 (e) Station 8 (h) Station 8

0.015 0.06 0.1
0.01 0.04
0.005 0.02 0.05
0 0 0
5 10 15 20 5 10 15 20 5 10 15

(c) Station 10 (f) Station 10 (i) Station 10

0.08
0.015 0.06 0.1
0.01 0.04
0.005 0.02 0.05
0 0 0
5 10 15 20 5 10 15 20 5 10 15
Time (s)

Figure 17
Water levels for e = (a–c) 0.05, (d–f) 0.3, and (g–i) 0.7 at station (a, d, g) 5, (b, e, h) 8, and (c, f, i) 10 for the model (solid black curves) and
observations (dashed grey curves). In each panel, the first and second waves are the incident and reflected waves, respectively
Vol. 168, (2011) Tsunami Model Validation: Toward Sediment Transport 2113

(a) 20

Elevation (m)
10

−10

−2000 −1500 −1000 −500 0 500 1000 1500 2000

Distance from Shoreline (m)
(b)
Peaked Peaked
Level (m)

5 Wave
Wave
Water

Broad
0
Wave
−5
0 20 40 60 80 100 120
Time (min)

Figure 18
a Observed (black circles) and predicted (dashed black curve) maximum water levels at Kuala Meurisi. The solid black curve is the combined
bathymetry and topography and the shaded grey area represents the range of maximum water levels predicted for the three simulations
conducted. b Offshore wave level shoaled to 17 m water depth

December 2004 tsunami (VATVANI et al., 2005a, b)
(Fig. 18b). Due to the coarse resolution of the deep
water model, the offshore water level variation was
available only in 35 m water depth. As the high
resolution nearshore bathymetry used in this study
extends only to 17 m water depth (Fig. 18a),
Green’s law (H * h1/4) is used to shoal the
predicted waves from 35 to 17 m water depth. This
shoaling produces approximately a 20% increase in
the wave height of the water level variation used to
force the model simulation. Changes in the wave
length are inherently included by forcing the model
with a time series, given that the wave period
remains relatively constant. However, no attempt
was made to simulate the change in wave shape due
to shoaling. The entire wave train, which is
composed of three large waves (i.e., two peaked
waves and a broad wave) followed by a series of
smaller waves, is simulated. This wave forcing is a
simplification of the actual tsunami as it is assumed
to be normally incident, and therefore neglects
waves propagating at angles to the coast, such as
trapped and scattered waves, and may not capture
accurately all wave reflections off nearby land
masses. Bed roughness is estimated using Manning’s
formulation with n = 0.025 and 0.032 off- and
onshore. A larger onshore roughness is used to take
into account the presence of vegetation.
3.8.2 Results
The model predicts the run-up elevation well, but
overpredicts the observed water levels closest to the
shoreline while underpredicting the water levels
further shoreward (Fig. 18a, compare dashed black
curve and open circles). The observed overprediction
near the shoreline may be owing to biased estimates
of the observed water levels, as no large trees
remained standing after the tsunami. Given the large
uncertainty in the model setup, including in the
offshore water level and bed roughness, the assump-
tion of alongshore uniformity, the use of bathymetry
taken after the tsunami, and the exclusion of local
obstacles such as houses and trees, the model predicts
the observations surprisingly well.
As a first step to quantify the effect of the
uncertainty on the maximum water level predictions,
two further simulations were conducted. The first was
forced using the unshoaled water level variation
predicted in 35 m water depth (i.e., an assumed lower
bound on the wave forcing). The second was forced
using a 40% increase in the water level from 35 m
2114 A. Apotsos et al.
water depth coupled with a 50% decrease in the on-
and offshore roughness (i.e., an assumed upper
bound). These simulations indicate that while the
uncertainty has an effect on the predictions, the
predictions from all three simulations fall within the
range of the observations (Fig. 18a, compare shaded
grey area and black circles). Future work will
examine in more detail the sensitivity of the flow
predictions to the input conditions.
4. Discussion
4.1. Solitary Waves as Tsunamis
Over the past few decades many analytical and
laboratory studies of tsunami propagation and inun-
dation have used solitary waves as proxies for
tsunami waves. It was first suggested by HAMMACK
(1973), SEGUR (1973), and HAMMACK and SEGUR
(1974, 1978a, b) that a positive initial surface
disturbance of arbitrary shape (i.e., an initial tsunami
wave generated in deep water) will eventually
devolve into a series of solitary waves if the
propagation distance is long enough. Based on this
observation and the fact that solitary waves are easily
reproducible in the laboratory, are defined by only
two parameters (i.e., Hi and d), and can be modeled in
the limited spatial scales of most laboratory facilities,
solitary waves have continued to be used extensively
to model tsunamis, including in 5 of the 7 bench-
marks examined in this study.
Recently, however, a growing number of authors
have questioned the relevance of solitary waves as
proxies for tsunamis (MADSEN et al., 2008; CONSTAN-
TIN and JOHNSON, 2008; CONSTANTIN, 2009). Using
geophysically realistic scales of nonlinearity and
dispersion, these studies demonstrate that the distance
necessary for earthquake-generated tsunami waves to
devolve into solitons can be several times the fetch of
the largest ocean basin. Further limiting the relevance
of solitary waves as proxies for tsunamis is the fact
that their nonlinearity and length are not independent.
Instead, the relative length of a solitary wave
decreases with increasing nonlinearity. Owing to the
limited spatial dimensions of most laboratory facil-
ities and the need to record measurable
Pure Appl. Geophys.
hydrodynamics, usually only highly nonlinear, and
thus relatively short, solitary waves are simulated.
For example, in the laboratory studies examined here,
the nonlinearity of the solitary waves varies from
0.05 to 0.67. However, the nonlinearity of actual
tsunamis observed away from the nearshore is
typically much smaller than this, and previous studies
(MADSEN et al., 2008) have suggested that solitary
waves with e [ 0.05 are not long waves, and are
more similar to wind waves than to tsunamis.
Furthermore, observations and eyewitness
accounts from several recent tsunamis indicate that
the largest wave does not always arrive first (e.g.,
PAPADOPOULOS et al., 2006; CHOOWONG et al., 2008)
and that the first wave can arrive as a leading
depression (TADEPALLI and SYNOLAKIS, 1996). Both of
these observations argue against the development of
solitons, which requires the largest wave to arrive
first and be of positive orientation. Some studies have
suggested that leading depression N-waves are a
better representation of tsunamis (TADEPALLI and
SYNOLAKIS, 1994, 1996). However, these waves are
difficult to generate in the laboratory and have not
been widely used in experimental studies. Further-
more, the use of single waves, whether of solitary or
N shape, will not produce wave–wave interactions,
which may be important for accurately predicting
both the maximum inundation distance and sediment
transport of tsunamis.
Unfortunately, the limited horizontal dimensions
of laboratory facilities and the difficulty of generating
more complex wave forms will limit the nonlinearity,
number, and shape of the waves that can be
generated. Therefore, until more realistic experiments
can be conducted or more detailed observations from
actual tsunamis recorded, these laboratory experi-
ments will remain an important aspect of validating
tsunami propagation and inundation models.
4.2. Scaling Up to Sediment Transport
Extending numerical models to include sediment
transport increases the complexity and uncertainty of
the model. For example, near bed fluid velocities and
the suspension of sediment are highly sensitive to the
bed roughness, which may vary significantly
Vol. 168, (2011) Tsunami Model Validation: Toward Sediment Transport
spatially. Calculating sediment transport also requires
the delineation of the sediment source characteristics
(i.e., location, grain-size distribution, thickness),
which are often poorly constrained, as well as the
use of transport formulations developed empirically
for lower velocity flows.
Unfortunately, no standardized benchmarks cur-
rently exist with which to verify model predictions of
tsunami-induced sediment transport, though preli-
minary work conducted as part of a workshop held in
Friday Harbor, Washington, in 2007 has shown some
initial success (HUNTINGTON et al., 2007). The devel-
opment of laboratory-scale benchmarks is not
straightforward because these benchmarks must
include and balance the effects of sediment suspen-
sion, advection, and settling on the appropriate time
and length scales. This is difficult given the limited
size of most laboratory facilities and the fact that
sediment becomes cohesive at small grain sizes. The
use of solitary waves, which dominate laboratory
studies, may not be appropriate because the flow
velocities and accelerations induced are likely to be
different than those induced by long tsunami waves.
Development of sediment transport benchmarks
based on field observations of sediment deposits
collected after a tsunami is hindered by the fact that
the initial conditions (i.e., the nearshore wave form,
sediment source characteristics) and onshore rough-
ness, especially in the presence of vegetation, are
typically poorly constrained. Furthermore, even good
agreement with observed water levels (i.e., Sect.
3.8.2), often the only hydrodynamic information
available, does not necessary imply good agreement
with the time-varying velocities, which dictate the
suspension and transport of sediment, that occurred
during the inundating tsunami.
The difficulties associated with modeling tsu-
nami-induced sediment transport not-withstanding,
the Kuala Meurisi simulation conducted in Sect. 3.8
was extended to include the suspension and transport
of sediment. First, the water column was subdivided
into ten vertical sigma layers (i.e., layers representing
a constant percentage of the water column), with
increasing resolution near the bed, to accurately
represent the bottom boundary layer. An erodable
sediment bed 5 m thick, with a mean grain size of
400 lm was assumed to exist over the entire model
2115
domain. While the erodable bed thickness at this site
is unknown, the maximum predicted scour is less
than 5 m, and therefore the onshore deposition of
sediment is not limited by the thickness used.
Sediment transport is calculated by solving the
conservation of mass equations within the hydrody-
namic model. Bedload and suspended load transport
are calculated following VAN RIJN (1993; 2007a, b)
combined with a k-e turbulence model. Bedload is
calculated based on a sediment mobility number and
the critical velocity for the initiation of motion.
Suspended load is calculated based on a near-bed
reference concentration and a vertical diffusion
coefficient determined from the output of the turbu-
lence model. Sediment fluxes between the water
column and the bed are calculated within each grid
cell based on the vertical diffusion of sediment owing
to turbulent mixing and deposition from settling, and
the bottom morphology is updated accordingly at
each time step (LESSER et al., 2004; VAN RIJN et al.,
2004). Sediment-induced density stratification owing
to high suspended sediment concentrations is inher-
ently accounted for within the turbulence model by
adjusting the fluid density to include the mass of the
suspended sediment. The settling velocity is deter-
mined based on the clear water settling velocity of
VAN RIJN (1993) combined with the effects of
hindered settling following RICHARDSON and ZAKI
(1954). The combined sediment transport/morpho-
logical change model has been shown to model well
many coastal and estuarine settings (LESSER et al.,
2004; GERRITSEN et al., 2007; VAN RIJN et al., 2007).
Using this simplified setup, the model predicts the
observed sediment deposit thickness generally within
a factor of 2, though locally the predictions can be off
by more than an order of magnitude (Fig. 19,
compare dashed grey curve with black circles). Both
model simulations and field measurements show
thicker deposits in local topographic lows and thinner
deposits or erosion on local topographic highs. Most
of the sediment is deposited by the second peaked
wave, although a smaller fraction is deposited by the
first peaked wave. Very little sediment was eroded or
deposited by the broad wave. The observed peak in
sediment thickness near the cliff is composed of fine
sediment and is not captured by the model which uses
only a single grain size. Given the high level of
2116
Thickness (m)
0.4
Deposit
0.2
0 200 400
Deposit thickness measured near Kuala Meurisi (black circles) and change in bed elevation predicted by the model (dashed grey curve)
uncertainty involved in the model setup, the model-
data agreement is remarkably good. More work needs
to be done to examine the details of the predicted
sediment transport and deposition to ensure that the
model is accurately capturing the appropriate pro-
cesses. However, the detailed analyses required are
beyond the scope of this paper, and will be the subject
of future work.
4.3. Model Applicability
Delft3D predicts well the general hydrodynamic
observations associated with tsunami nearshore prop-
agation and inundation measured under a wide range
of conditions. This study, combined with previous
studies (VATVANI et al., 2005a, b; GELFENBAUM et al.,
2007) that showed Delft3D models well the deep-
water propagation and inundation of the 26 December
2004 Indian Ocean tsunami near Banda Aceh,
demonstrates that Delft3D is applicable for modeling
long, non-dispersive tsunamis from the source to the
maximum extent of inundation.
As Delft3D is based on the NLSWEs, which are
derived using the assumption of hydrostatic pressure,
the model does not include dispersive terms and is
best suited to tsunamis where the wavelength is long
compared to the water depth. The model-data com-
parisons presented here support this, with the model
predicting most accurately the observations from
benchmarks where the wave is long (i.e., the analyt-
ical solution over a planar beach) and where
dispersion is relatively less important (i.e., Trial05
from the piece-wise linear experiment). Therefore,
Delft3D is likely appropriate for modeling most
earthquake-induced tsunamis, which normally have
very long wavelengths and experience negligible
frequency dispersion over typical propagation
distances.
A. Apotsos et al. Pure Appl. Geophys.
600 800 1000 1200 1400 1600 1800 2000
Distance from Shoreline (m)
Figure 19
As stated previously, comparisons with laboratory
observations offer only qualified validation of model
applicability. Therefore, while the model predictions
are in reasonable agreement with benchmarks in
which dispersion is important (i.e., highly nonlinear
solitary waves), this does not necessarily imply the
model is appropriate for modeling highly dispersive
tsunamis. Tsunamis generated by landslides, volcanic
eruptions, and meteoroid impacts can have wave-
lengths of the same order of magnitude as the water
depth, and frequency dispersion can be important
even near the source. A previous study (LYNETT et al.,
2003) found large differences between offshore wave
height predictions from models based on the NLS-
WEs and on the Boussinesq equations, which allow
for weak frequency dispersion, for the landslide-
generated component of the 1998 Papua New Guinea
tsunami. Therefore it is unclear how accurately
Delft3D will model the deep-water propagation of
highly dispersive waves such as those generated by a
landslide.
The good model-data agreement demonstrates the
numerical solution method employed is appropriate
for modeling the general characteristics (i.e., wave
height, run-up, velocities) of both breaking and non-
breaking waves. The advantage of this numerical
solution technique over many Boussinesq models is
that it does not require a breaking criterion, but
inherently captures the loss of energy associated with
wave breaking. Furthermore, the accuracy of the
predictions onshore of breaking does not appear to
decrease for waves with higher initial nonlinearity
(e.g., Fig. 5d), which may suggest depth-limited
breaking masks the effects of frequency dispersion
onshore of breaking. A previous study (LYNETT et al.,
2003) found similar nearshore wave height predic-
tions using both an NLSWE and a Boussinesq model
for the landslide-generated component of the 1998
Vol. 168, (2011) Tsunami Model Validation: Toward Sediment Transport
Papua New Guinea tsunami even though the model
predictions further offshore were significantly differ-
ent. It was suggested that the similarity in the
nearshore wave height predictions was owing to
depth-limited breaking. While this may imply that
Delft3D can be used to model the run-up and
inundation of breaking, dispersive waves, under
exactly what conditions the NLSWEs accurately
represent the hydrodynamics of dispersive tsunamis
onshore of breaking remains to be determined.
While the model predicts well the overall dissi-
pation of wave energy owing to breaking, the model
does not predict local breaking dynamics such as the
generation of turbulence or vertical flow accelera-
tions. Even for waves that are non-dispersive in deep
water, dispersion becomes important close to break-
ing where it balances the increasing nonlinearity
associated with shoaling and delays breaking (CONST-
ANTIN and JOHNSON, 2008). As the model does not
include dispersion, it will predict that waves break
further offshore than physically realistic. Therefore,
while the model may predict well the water levels and
velocities before and after breaking, as well as the
maximum run-up elevations of breaking waves, it is
not appropriate for simulating wave characteristics
near the breakpoint. Furthermore, the model may
underpredict the run-up and inundation of waves
close to breaking, as the lack of dispersion will cause
these otherwise non-breaking waves to break and
dissipate energy.
The generally good agreement with the observed
water levels and flow velocities suggests extension
of the model to include sediment transport may be
appropriate for long, non-breaking waves generated
by earthquakes. However, extension of the model
may not be appropriate for dispersive or breaking
waves. In the case of dispersive waves, the wave
shape, and thus the cross-shore distribution of
velocity, is poorly predicted, and the model may
misrepresent the distribution of the bed stress. For
breaking waves, the location and magnitude of
wave-breaking-generated turbulence, which the
model does not predict for long waves, will affect
the patterns of sediment suspension, transport, and
deposition.
2117
5. Conclusions
The model predicts well the hydrodynamic
observations (i.e., water levels, velocities, and inun-
dation distances) from seven analytical and
laboratory experiments and the maximum water lev-
els observed near Kuala Meurisi, Sumatra, following
the 26 December 2004 tsunami. Model predictions
agree best with experiments where dispersion is less
important. The model predicts well the total decrease
in wave height and the speed of breaking waves.
However, the model does not reproduce the observed
shoaling for some waves or the shape of highly
nonlinear waves in deep water. These results indicate
the model is appropriate for simulating tsunami
propagation and inundation for both breaking and
non-breaking long waves (i.e., most earthquake-
induced tsunamis) but may not be appropriate for
modeling highly dispersive tsunamis, such as those
generated by landslides or impacts, especially over
long distances.
Extension of the model to include sediment
transport is shown to produce reasonable results for
long, non-breaking waves, but may not be appropriate
in the case of breaking or dispersive waves. While
including sediment transport increases the complexity
of and uncertainty in the model, the model predicts
the sediment deposition at Kuala Meurisi generally
within a factor of two. These results suggest that the
coupling of validated tsunami-induced sediment
transport models with paleo-tsunami deposits could
aid in the improvement of local tsunami hazards
assessments.
Acknowledgements
This research was funded by a USGS Mendenhall
Post-Doctoral Fellowship and the USGS Coastal
and Marine Geology Program. Reviews provided
by Eric Geist, Pat Lynett, Utku Kânoğlu, and 2
anonymous reviewers greatly improved this manu-
script. Edwin Elias is thanked for his help and
guidance concerning the details of the numerical
model.
2118 A. Apotsos et al. Pure Appl. Geophys.

REFERENCES HAMMACK, J. (1973), A note on tsunamis: Their generation and

propagation in an ocean of uniform depth, J. Fluid Mech., 60(4),
769-799.
ARCAS, D. and AITOV, V. (2006), Sumatra tsunami: lessons from
HAMMACK, J. and SEGUR, H. (1974), The Korteweg-deVries equation
modeling, Surv. Geophys., 27, 679-705.
and water waves, part 2: Comparison with experiments, J. Fluid
ARCEMENT, G. and SCHNEIDER, V. (1990), Guide for selecting
Mech., 65(2), 289-314.
Manning’s roughness coefficients for natural channels and flood
HAMMACK, J. and SEGUR, H. (1978a), The Korteweg-deVries equa-
plains, Water Supply Pap., U.S. Geol. Surv., Reston, VA.
tion and water waves, part 3: Oscillatory waves, J. Fluid Mech.,
ATWATER, B.F. (1987), Evidence for great Holocene earthquakes
84(2), 337-358.
along the outer coast of Washington State, Science, 236,
HAMMACK, J. and SEGUR, H. (1978b), Modeling criteria for long
942-944.
waves, J. Fluid Mech., 84(2), 359-373.
BALDOCK, T.E., COX, D., MADDUX, T., KILLIAN, J., and FAYLER, L.
HIBBERD, S. and PEREGRINE, D. (1979), Surf and run-up on a beach:
(2008), Kinematics of breaking tsunami wave fronts: A dataset
a uniform bore, J. Fluid Mech., 95, 323-345.
from large scale laboratory experiments, Coastal Eng., doi:
HUNTINGTON, K., BOURGEOIS, J., GELFENBAUM, G., LYNETT, P., JAFFE,
10.1016/j.coastaleng.2008.10.011.
B., YEH, H., and WEISS, R. (2007), Sandy signs of a tsunami’s
BRIGGS, M.J., SYNOLAKIS, C.E., HARKINS, G.S., and GREEN, D.R.
onshore depth and speed, EOS Trans., 88(52), 577-584.
(1995), Laboratory experiments of tsunami run-up on a circular
JAFFE, B. and GELFENBAUM, G. (2007), A simple model for calcu-
island, Pure Appl. Geophys., 144, 569-593.
lating tsunami flow speed from tsunami deposits, Sediment.
BRIGGS, M.J., SYNOLAKIS, C.E., KANOGLU, U., and GREEN, D.R.
Geol., 200, 347-361.
(1996), Benchmark Problem #3: Runup of solitary waves on a
JAFFE, B., BORRERO, J., PRASETYA, G., PETERS, R., MCADOO, B.,
vertical wall, ‘‘Long-wave Runup Models,’’ International
GELFENBAUM, G., MORTON, R., RUGGIERO, P., HIGMAN, B., DEN-
Workshop on Long Wave Modeling of Tsunami Runup, Friday
GLER, L., HIDAYAT, R., KINGSLEY, E., KONGKO, W., LUKIJANTO,
Harbor, San Juan Island, WA, September 12-16, 1995.
MOORE, A., TITOV, V., and YULIANTO, E. (2006), Northwest
BROCCHINI, M. and PEREGRINE, D. (1996), Integral flow properties of
Sumatra and offshore islands field survey after the December
the swash zone and averaging, J. Fluid Mech., 317, 241-273.
2004 Indian Ocean tsunami, Earthquake Spectra, 22, 105-135.
CARRIER, G.F. and GREENSPAN, H.P. (1958), Water waves of finite
KÁNOĞLU, U. (2004), Nonlinear evolution and runup-rundown of
amplitude on a sloping beach, J. Fluid Mech., 4, 97-109.
long waves over a sloping beach, J. Fluid Mech., 513, 363-372.
CARRIER, G.F., WU, T.T., and YEH, H. (2003), Tsunami run-up and
KÁNOĞLU, U. and SYNOLAKIS, C.E. (1998), Long wave runup on
draw-down on a plane beach, J. Fluid Mech., 475, 79-99.
piecewise linear topographies, J. Fluid Mech., 374, 1-28.
CHOOWONG, M., MURAKOSHI, N., HISADA, K., CHARUSIRI, P.,
LIU, P.L.-F., CHO, Y.S., BRIGGS, M.J., KANOGLU, U., and SYNOLAKIS,
CHAROENTITIRAT, T., CHUTAKOSITKANON, V., JANKAEW, K.,
C.E. (1995), Run-up of solitary waves on a circular island, J.
KANJANAPAYONT, P., and PHANTUWONGRAJ, S. (2008), 2004 Indian
Fluid. Mech., 302, 259-285.
Ocean tsunami inflow and outflow at Phuket, Thailand, Mar.
LIU, P.L.-F., YEH, H., and SYNOLAKIS, C.E. (ed.) (2008), Advanced
Geo., 248, 179-192.
numerical models for simulating tsunami waves and run-up,
CONSTANTIN, A. (2009), On the relevance of soliton theory to tsu-
Advances in Coastal and Ocean Engineering, 10.
nami modeling, Wave Motion, 6, 420-426.
LESSER, G.R., ROELVINK, J.A., VAN KESTER, J.A.T.M., and STELLING,
CONSTANTIN, A. and JOHNSON, R.S. (2008), Propagation of very long
G.S. (2004), Development and validation of a three-dimensional
waves, with vorticity, over variable depth, with applications to
morphological model, Coastal Eng., 51, 883-915.
tsunamis, Fluid Dyn. Res., 40, 175-211.
LYNETT, P., BORRERO, J., LIU, P., and SYNOLAKIS, C.E. (2003), Field
FRITZ, H., BORRERO, J., SYNOLAKIS, C., and YOO, J. (2006), 2004
survey and numerical simulations: A review of the 1998 Papua
Indian Ocean flow velocity measurements from survivor videos,
New Guinea Tsunami, Pure Appl. Geophys., 160, 2119-2146.
Geophys. Res. Lett., 33, L24605.
MADSEN, P., FUHRMAN, D., and SCHAFFER, H. (2008), On the solitary
GELFENBAUM, G., VATVANI, D., JAFFE, B., and DEKKER, F. (2007),
wave paradigm for tsunamis, J. Geophys. Res., 113, C12012, doi:
Tsunami inundation and sediment transport in the vicinity of
10.1029/2008JC004932.
coastal mangrove forest, Conference Proceedings Coastal Sedi-
NANAYAMA, F., FURUKAWA, R., SHIGENO, K., MAKINO, A., SOEDA, Y.,
ment ‘07, doi: 10.1061/40926(239)86.
and IGARASHI, Y. (2007), Nine unusually large tsunami deposits
GERRITSEN, H., DE GOEDE, E.D., PLATZEK, F.W., GENSEBERGER, M.,
from the past 4000 years at Kiritappu marsh along the southern
VAN KESTER, J.A.TH.M., and UITTENBOGAARD, R.E. (2007), Vali-
Kuril Trench, Sediment. Geol., 200, 275-294.
dation document for Delft3D-Flow, Rep. X0356, Delft
PAPADOPOULOS, G., CAPUTO, R., MCADOO, B., PAVLIDES, S., KA-
Hydraulics, Delft, The Netherlands.
RATATHIS, V., FOKAEFS, A., ORFANOGIANNAKI, K., and VALKANIOTIS,
GONZALEZ, F., BERNARD, E., MEINIG, C., EBLE, M., MOFJELD, H., and
S. (2006), The large tsunami of 26 December 2004: Field
STALIN, S. (2005), The NTHMP Tsunameter network, Nat.
observations and eyewitnesses accounts from Sri Lanka, Mal-
Hazards, 35, 25-29.
dives Is. and Thailand, Earth, Planets, Space, 58, 233-241.
GONZÁLEZ, F.I., GEIST, E.L., JAFFE, B., KÂNOĞLU, U., MOFJELD, H.,
RICHARDSON, J.F. and ZAKI, W.N. (1954), Sedimentation and fluid-
SYNOLAKIS, C.E., TITOV, V.V., ARCAS, D., BELLOMO, D., CARLTON,
ization: Part I, Trans. Inst. Chem. Eng., 32, 35-50.
D., HORNING, T., JOHNSON, J., NEWMAN, J., PARSONS, T., PETERS, R.,
SEGUR, H. (1973), The Korteweg-deVries equation and water
PETERSON, C., PRIEST, G., VENTURATO, A., WEBER, J., WONG, F.,
waves, part 1: Solutions of the equations, J. Fluid Mech., 59(4),
and YALCINER, A., (2009), Tsunami pilot study working group
721-735.
‘‘Probabilistic tsunami hazard assessment at Seaside, Oregon for
SMITH, D.E., FOSTER, I.D.L., LONG, D., and SHI, S. (2007), Recon-
near- and far-field seismic sources,’’ J. Geophys. Res., 114,
structing the pattern and depth of flow onshore in a paleo-
C11023, doi:10.1029/2008JC005132.
Vol. 168, (2011) Tsunami Model Validation: Toward Sediment Transport
tsunami from associated deposits, Sediment. Geol., 200,
362-371.
STELLING, G.S. and DUINMEIJER, S.P.A. (2003), A numerical method
for every Froude number in shallow water flows, including large
scale inundations, Int. J. Num. Meth. in Fluids, 43, 1329-1354.
STELLING, G.S. and LEENDERTSE, J.J. (1991), Approximation of
convective processes by cyclic ACI methods. In: Proceedings of
2nd ASCE Conference on Estuarine and Coastal Modeling,
Tampa, USA, 771–782.
STELLING, G.S. and VAN KESTER, J.A.T.M. (1994), On the approxi-
mation of horizontal gradients in sigma coordinates for
bathymetry with steep bottom slopes, Int. J. Num. Meth. in
Fluids, 18, 915-935.
SVENDSEN, I.A. and MADSEN, P.A. (1984), A turbulent bore on a
beach, J. Fluid Mech., 148, 73-96.
SWILGER, D.T. (2009), Laboratory study investigating the three
dimensional turbulence and kinematic properties associated with
a breaking solitary wave, Master Thesis, Texas A&M.
SYNOLAKIS, C.E. (1987), The run-up of solitary waves, J. Fluid
Mech., 185, 523-545.
SYNOLAKIS, C.E., BERNARD, E.N., TITOV, V.V., KANOGLU, U., and
GONZALEZ, F. (2007), Standards, criteria, and procedures for
NOAA evaluation of tsunami numerical models, NOAA OAR
Special Report, Contribution No 3053, NOAA/OAR/PMEL,
Seattle, Washington.
SYNOLAKIS, C.E., BERNARD, E.N., TITOV, V.V., KANOGLU, U., and
GONZALEZ, F. (2008), Validation and verification of tsunami
numerical models, Pure Appl. Geophys., 165, 2197-2228.
TADEPALLI, S. and SYNOLAKIS, C.E. (1994), The run-up of N-waves
on sloping beaches, Proc. R. Soc. Lond., 445, 99-112.
TADEPALLI, S. and SYNOLAKIS, C.E. (1996), Model for the leading
waves of tsunamis, Phys. Rev. Lett., 2141-2144.
TANG, L., TITOV, V., and CHAMBERLIN, C. (2009), Development,
testing, and application of site-specific tsunami inundation
models for real-time forecasting, J. Geophys. Res., 114, C12025,
doi:10.1029/2009JC005476.
TITOV, V. and SYNOLAKIS, C.E. (1997) Numerical modeling of 2-D
and 3-D long wave runup using VTSC-2 and VTSC-3. In Long
wave runup models (ed. H. Yeh et al.), World Scientific,
242-248.
(Received January 21, 2010, revised December 8, 2010, accepted December 19, 2010, Published online March 4, 2011)
2119
TITOV, V., GONZALEZ, F., BERNARD, E., EBLE, M., MOFJELD, H.,
NEWMAN, J., and VENTURATO, A., (2005), Real-time tsunami
forecasting, Natural Hazards, 35(1), 45-58.
U.S. Agency for International Development (USAID) (2005), Fact
sheet, July 7. http://www.usaid.gov/our_work/humanitarian_
assistance/disaster_assistance/countries/Indian_ocean/fy2005/
indianocean_et_fs39-07-07-2005.pdf.
VAN RIJN, L.C. (1993), Principles of sediment transport in rivers,
estuaries and coastal seas, Aqua Publications, Amsterdam.
VAN RIJN, L.C. (2007a), Unified view of sediment transport by
currents and waves. I: Initiation of motion, bed roughness, and
bed-load transport, J. Hydraul. Eng., 133, 649-667.
VAN RIJN, L.C. (2007b), Unified view of sediment transport by
currents and waves. II: Suspended transport, J. Hydraul. Eng.,
133, 668-689.
VAN RIJN, L.C., WALSTRA, D.J.R., and VAN ORMONDT, M. (2004),
‘‘Description of TRANSPOR 2004 (TR2004) and implementation
in DELFT3D-online,’’ Rep. Z3748, Delft Hydraulics, Delft, The
Netherlands.
VAN RIJN, L.C., WALSTRA, D., and VAN ORMONDT, M. (2007), Unified
view of sediment transport by currents and waves. IV: Applica-
tion of morphodynamic model, J. Hydraul. Eng., 133, 776-793.
VATVANI, D., SCHRAMA, E., and VAN KESTER, J. (2005a), Hindcast of
tsunami flooding in Aceh-Sumatra, Proc. of the 5th International
Symposium on Ocean Wave Measurement and Analysis, Madrid,
Spain.
VATVANI, D., BOON, J., and RAMANAMURTY, P.V. (2005b), Flood risk
due to tsunami and tropical cyclones and the effect of tsunami
excitations on tsunami propagations, Proc. of the IAEA Work-
shop on External Flooding Hazards, Tamil Nadu, India.
WEI, Y., BERNARD, E., TANG, L., WEISS, R., TITOV, V., MOORE, C.,
SPILLANE, M., HOPKINS, M., and KÁNOĞLU, U. (2008), Real-time
experimental forecast of the Peruvian tsunami of August 2007 for
U.S. coastlines, Geophys. Res. Letters, 35, L04609.
YEH, H., LIU, P., and SYNOLAKIS, C.E., (ed.) (1996), Proc. Int.
Symposium, Friday Harbor, USA, 12–17 September 1995, Long-
Wave Rump Models, World Sci., Singapore.
ZIJLEMA, M. and STELLING, G.S. (2008), Efficient computation of
surf zone waves using the nonlinear shallow water equations
with non-hydrostatic pressure, Coastal Eng., 55, 780-790.